A new chip will teach robots to anticipate movement.

Stand in the way of an industrial robot - stupidthe idea. These cars move quickly and accurately, but they are blind and stupid, so they will gladly break your arm and pierce the skull if you find yourself in their path. Therefore, the testers had to take courage and try this trick with a robotic manipulator in Realtime Robotics, a startup located in the area of ​​the Boston seaport. The tester reached forward to intercept the movement of the robot, when he grabbed the part from the table and sent it to the box.

Fortunately, the robot paused, deftly circled the outstretched arm and neatly folded the item into the box. There will be no broken hands today.

Such graceful adaptability may beincredibly useful for robotics. Some robots may work with humans, but they are usually thin, inaccurate, and used sparingly. The most powerful and capable industrial machines still have to work in very precisely controlled conditions, away from soft, brittle people.

Why are industrial robots dangerous to humans?

Some companies are trying to find ways aroundthis problem. Some are testing sensors that will stop powerful robots on their way if they find an obstacle. Realtime Robotics goes further, providing the robots with the low level intellectual abilities necessary to move around the real world. People and animals take this physical awareness for granted whenever they raise an arm or a leg.

Multiple rooms Realtime Roboticsindustrial robotic arms are testing the capabilities of the new chip developed by the company. When connected to three-dimensional sensors, such a chip allows the machines to quickly consider a number of different actions, essentially “imagining” the result, before choosing the most suitable for the task.

“The main problem is that robots are very stupid,” says George Conidaris, founder and chief robotics engineer of Realtime Robotics. "We have this basic motor skills, but robots do not."

Movement planning is deceptively difficult for a robot, in part because each joint adds an extra dimension to the calculations that need to be performed.

A company chip performs mathematical calculations.using a relatively simple motion planning algorithm developed by Conidaris and his colleagues while he was still studying at Duke University. With parallel computations, a dedicated chip can perform them in more than 10,000 faster than a conventional computer chip, while consuming less power.

This is part of a broader trend. Advances in software and hardware gradually lead to an increase in the IQ of robots, thereby opening up new opportunities for use. Smarter robots can sit on the production line next to a person - understanding how to pick up objects, regardless of their location and without accidentally harming anyone. This can speed up automation in many industries.

Improved motion planning will be crucial for the future of robotics.

"Any robot that intentionally moves forachieving goals, should be better aware of how it should move, ”scientists say. "The problem is that movement planning is slower if the environment is tangled, especially when the robot has many degrees of freedom."

There is another big potential use.this technology: self-driving cars. Just as a robot needs to plan its movement if it wants to avoid a collision, so the self-driving car needs to choose the safest route around the obstacles. Realtime Robotics is already developing a special chip for this. Conidaris says that this should allow self-driving cars to quickly adapt to the difficulties on the road, possibly making them safe.

It all depends on whether manufacturers of manipulators decide to use Realtime Robotics technology. Some are already testing it. The ultimate success will depend on the integration.

The hour is not far off when robots will stop disturbing people and become safe for them. Do you agree? Tell us in our chat in Telegram.